Electric Field Cycling Behavior of Ferroelectric Hafnium Oxide - ACS

Nov 3, 2014 - Warren , W. L.; Vanheusden , K.; Dimos , D.; Pike , G. E.; Tuttle , B. A. ...... Kuen-Yi Chen , Ka-Lip Chu , Pin-Hsuan Chen , Yung-Hsien...
0 downloads 0 Views 1MB Size
Subscriber access provided by ONDOKUZ MAYIS UNIVERSITESI

Article

Electric Field Cycling Behavior of Ferroelectric Hafnium Oxide Tony Schenk, Uwe Schroeder, Milan Peši#, Mihaela Popovici, Yuriy V. Pershin, and Thomas Mikolajick ACS Appl. Mater. Interfaces, Just Accepted Manuscript • DOI: 10.1021/am504837r • Publication Date (Web): 03 Nov 2014 Downloaded from http://pubs.acs.org on November 7, 2014

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

ACS Applied Materials & Interfaces is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Electric Field Cycling Behavior of Ferroelectric Hafnium Oxide Tony Schenk* ,†, Uwe Schroeder†, Milan Pešić†, Mihaela Popovici‡, Yuriy V. Pershin§, Thomas Mikolajick†,∥ †

NaMLab gGmbH, Noethnitzer Str. 64, D-01187 Dresden, Germany



Imec, Kapeldreef 75, B-3001 Leuven, Belgium

§

Department of Physics and Astronomy, and the University of South Carolina Nanocenter,

University of South Carolina, Columbia, South Carolina 29208, USA ∥

Chair of Nanoelectronic Materials, TU Dresden, D-01062 Dresden, Germany

ABSTRACT HfO2 based ferroelectrics are lead-free, simply binary oxides with non-perovskite structure and low permittivity. It just recently started attracting attention of theoretical groups and also in the fields of ferroelectric memories and electrostatic supercapacitors. A modified approach of harmonic analysis is introduced for temperature dependent studies of the field cycling behavior and the underlying defect mechanisms. Activation energies for wake-up and fatigue are extracted. Notably, all values were in the order of 100 meV, which is one order of magnitude lower than for conventional ferroelectrics like lead zirconate titanate (PZT). This difference is mainly attributed to result from the one to two magnitudes higher electric fields used for cycling and different surface to volume ratios between the 10 nm thin films in this study

ACS Paragon Plus Environment

1

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 25

and the bulk samples of former measurements or simulations. Moreover, a new, analog-like split-up effect of switching peaks by field cycling is discovered and explained by a network model based on memcapacitive behavior as a result of defect redistribution.

KEYWORDS Hafnium Oxide, Ferroelectrics, Harmonic Analysis, Fatigue, Wake-up, Field Cycling Behavior

INTRODUCTION Hafnium oxide was recently reported to exhibit ferroelectric behavior.1 From the material point of view, HfO2 based ferroelectrics are of special interest because they are lead-free, simple binary oxides and exhibit a non-perovskite structure with a comparably low relative permittivity in the range of ~ 20 – 301. Thus, first simulation efforts with focus on ferroelectricity in HfO2 were started.2,3 Moreover, a recent simulation-based material design search indicated that there are further (anti-) ferroelectric compounds to be discovered4 and motivated a theoretical work on antiferroelectricity in the chemically very similar ZrO25. Nonetheless, this marks just a starting point. Many questions concerning the tuning of the ferroelectric properties and the role of defects are studied far less extensively than e.g. for lead zirconate titanate (PZT). Thus, the transfer of promising analysis methods is the first step for deepening the understanding of this material system. From the application point of view, HfO2 based ferroelectrics attracted attention especially in the field of ferroelectric memories1,6,7 and recently also for electrostatic supercapacitors8. Since 2007, HfO2 is the state-of-the-art gate dielectric in complementary metaloxide-semiconductor technology10 and therefore, also a promising candidate to overcome

ACS Paragon Plus Environment

2

Page 3 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

integration and scaling issues associated with conventional ferroelectrics.11 Ferroelectric field effect transistors in 28 nm technology12 evidenced this potential already in 2012.

RESULTS AND DISCUSSION The present work is dedicated to the field cycling behavior of Sr:HfO2, which showed ferroelectric hystereses with remanent polarizations up to 23µC/cm² and the highest coercive fields (~ 2 MV/cm) of all tested dopants so far13. Besides, the conditioning behavior of a pristine sample (wake-up) was shown: Two distinct current peaks merge during cycling causing a constricted polarization hysteresis to become more ideal and open. Such constricted hystereses were shown by Carl and Hardtl14 for doped PZT and attributed to internal bias fields. These fields were assumed to result from incorporated defects, which either pin domains walls or form anisotropic centers favoring one direction of the spontaneous polarization.14 Further possible roots for constrictions and deformations of hystereses were reviewed elsewhere.15 This paper is focused on: 1) A new analog split-up phenomenon of ferroelectric switching peaks in the transient current response which was modelled as a result of defect redistribution. 2) Up to now, the first temperature dependent measurements to determine activation energies for the three defect related effects wake-up, fatigue and the aforementioned split-up phenomenon. A new convenient and time saving approach of harmonic analysis was introduced for this purpose.

“Wake-up” refers to the conditioning effect of a pristine sample, as it was reported also for Y16 or Si doped HfO212,17,18 and for PZT19: The pinched hysteresis of a pristine sample opens as the

ACS Paragon Plus Environment

3

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 25

electric field is progressively cycled up and down. In the observed current response, this is reflected in the merging of two or more distinct current maxima into one single peak on each side of the electric field, as can be seen from Figure 1a).

Figure 1. a) Wake-up effect: The constricted hysteresis of a pristine sample opens after field cycling and two distinct current peaks merge into one single peak. b) Current peak split-ups occur directly at the values of the cycling field amplitude (constant no. of cycles). c) Split-up of multiple peaks by subsequent field cycling with different field amplitudes in descending order. The split-up at 1.45 MV/cm is already vanished in the first quadrant (red line). All experiments were performed with f = 1000 Hz. The corresponding P-E hystereses are shown as insets with the E- and P-axes scaled from –4 MV/cm to 4 MV/cm and –30 µC/cm2 to 30 µC/cm2, respectively.

ACS Paragon Plus Environment

4

Page 5 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Applying an electric field that cycles the ferroelectric in a subloop of the former wake-up procedure, an increasingly constricted hysteresis can be observed. The corresponding curves of the current I vs. the electric field E, then, show a progressing peak split-up right at the voltage of the cycling field (see Figure 1b)). The more subcycles are applied, the more pronounced the split-up in the I-E and constriction in P-E graph becomes (see Figure S1 in the supporting information). The procedure of re-merging the peaks by the cycling with the original, high field amplitude of 3.75 MV/cm shows a clear trend: The higher the subcycling amplitude, the more cycles are required to both, split-up the corresponding two peaks during subcycling and later on to re-merge them again with 3.75 MV/cm amplitude. As mentioned before, constricted hystereses and their depinching by a higher field amplitude were reported earlier also for PZT. But, to the authors’ knowledge, a pinching by field cycling with lower amplitude has not been reported, yet. In addition to this new observation, the split-up of multiple peaks is possible as shown in Figure 1c). For this purpose, the desired cycling amplitudes have to be applied in descending order to not re-merge previously split-up peaks. This peak split-up appears to be possible in an analogous way and in any arbitrary configuration only depending on the appropriate selection of voltages and their according number of split-up cycles. In the following, harmonic analysis20,21 was applied to study this effect and also the defect related phenomena of polarization fatigue and wake-up. Morozov and Damjanovic used a setup with a lock-in amplifier to analyze amplitudes and phases up to the 9th harmonic. They found a phase jump in the 3rd harmonic during the transition from a constricted to an open hysteresis. Exploiting such features that mark a clear point in time makes this method superior to simply arguing with the P-E or I-E curves. Moreover, comparisons to predictions of hystereses models

ACS Paragon Plus Environment

5

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 25

like the Preisach model22 can be drawn (see Supporting Information S3 and S4). Nonetheless, the method is not limited by the conformity of the hystereses to any model prediction. If a certain feature such as a phase jump can be related to the change in the hysteresis shape, it can be used to mark a clear point in time for the transition from e.g. constricted to non-constricted. Measurements with different frequencies allow distinguishing whether the cycling time or the no. of switching processes is the figure of interest for the underlying mechanism. However, this lock-in technique suffers one major drawback: A certain number of inputchannels is needed to monitor all harmonics at one stretch. Measuring only one harmonic at a time multiplies the measurement time tremendously. Moreover, no P-E or I-E curves are obtained. The alternative presented in this work is based on a commercial test system for ferroelectrics and its “Fatigue Measurement” environment. It allows automatically acquiring as much dynamic hysteresis measurements (P-E and I-E curves) per decade of field cycles as regarded necessary. Afterwards, the measured hystereses are subjected to a Fourier series expansion to calculate the corresponding amplitude and phase values:

P(t ) = P0 + P1 ⋅ sin (ωt + ϕ1 ) + P2 ⋅ sin ( ωt + ϕ2 ) + ... , (1) where Pk represents the amplitude and ϕk the phase of the k-th harmonic. This convenient approach is now applied to study, first, the wake-up effect and later on also fatigue and the split-up phenomenon described at the beginning. The wake-up process slows down as the temperature T is decreased as shown exemplarily in Figure 2. At 171 K and 10 kHz, about 100 cycles were necessary to no longer see two switching peaks (i.e. a pronounced kink around the coercive field), whereas at 136 K, it took about 1000 cycles. It can be seen that it is difficult to assess the exact point where the two peaks are merged into one peak. As mentioned before, the harmonic analysis provides a potentially more precise quantitative study of the T-

ACS Paragon Plus Environment

6

Page 7 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

dependence. Moreover, additional features are observed: The switching peaks become broader for lower temperatures or higher frequencies and the remanent and maximum polarization decreases with increasing frequency. This indicates that the maximum voltage used here, was not high enough to sufficiently saturate all hystereses. Dielectric breakdown limited the applicable electric fields.

Figure 2. I-E hystereses for 100 Hz and 10 kHz at 136 K and 171 K . The corresponding P-E hystereses are shown as insets with the E- and P-axes scaled from –4 MV/cm to 4 MV/cm and – 30 µC/cm2 to 30 µC/cm2, respectively.

Figure 3a) exemplarily shows the evolution of amplitude and phase of the first ten harmonics against the number of fatigue cycles cycling time for 10 kHz at 171 K. A phase jump and an accompanying minimum in amplitude can be seen for the 7th and 9th harmonic. The times of the

ACS Paragon Plus Environment

7

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 25

phase jumps were extracted for all temperatures and are plotted into the Arrhenius-plot in Figure 3b). Activation energies EA of 105 meV and 126 meV were extracted for 7th and 9th harmonic, respectively. Since the harmonics are only a mathematical representation of the hysteresis shape, the difference might be surprising, but is consistent also for all further observations in this work. This could be interpreted as follows: From Figure 3, it can be seen that phase jumps of higher harmonics occured later in time. Thus, different activation energies could be explained by different mechanisms dominating the hysteresis change at the beginning and at the end of the wake-up process. Moreover, a clear frequency dependence is observed, which can, at least partially, be attributed the frequency dependent change in the switching characteristics as discussed in more detail in section S5 of the Supporting Information. At 1 kHz, only 98 meV and 99 meV are extracted for 7th and 9th harmonic, respectively. For 100 Hz, the values of 84 meV and 88 meV are even lower. In contrast to Morozov and Damjanovic20,21, no pure time dependence was observed. Thus, also the number of polarization reversals is important. Quite likely, Morozov and Damjanovic saw only a phase jump in the 3rd harmonic since their observation time was too short to see jumps in higher odd harmonics at the temperatures used. There is no particular reason why only the 3rd harmonic should be affected by the pinching/depinching of a hysteresis.15

ACS Paragon Plus Environment

8

Page 9 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Figure 3. a) Example of the evolution of amplitude (normalized to the respective initial value) and phase against the number of fatigue cycles for f = 10000 Hz and T = 171 K. Jumps in the phase of the 7th and 9th harmonic occur subsequently and are accompanied by a minimum in the amplitude. b) Arrhenius-plot of the cycling time at which the phase jumps of the respective harmonic was observed for f = 10000 Hz. Data points were extracted from a) as shown and similarly for the other temperatures between 136 K and 298 K.

The effect of polarization fatigue, as it was formerly shown for Sr:HfO2 in ref. 13, was studied by plotting the cycling time at which the remanent polarizations reached its maximum in an Arrhenius-plot (Figure S5). Activation energies of 103 meV, 56 meV and 48 meV were extracted for 10 kHz, 1 kHz and 100 Hz, respectively. Besides the lower absolute values of EA, the more pronounced frequency dependence gives rise to the assumption that different mechanisms are responsible/dominant for fatigue and wake-up (compare section S6 in the Supporting Information). The novel subcycling induced split-up effect or more precisely the re-merging after split-up was studied analogously to the wake-up effect by harmonic analysis. First, the samples were subjected to a treatment consisting of 103 cycles of 3.75 MV/cm amplitude to ensure a proper wake-up and 105 subcycles to induce a split-up into two distinct current peaks. Independent of

ACS Paragon Plus Environment

9

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 25

the frequency used for merging, 1 kHz was used for this pre-treatment. To prevent an imprint23, a depolarization sequence consisting of ten cycles with linearly decreasing amplitude from 3.75 MV/cm to 0 MV/cm was applied. Afterwards, the samples were cooled down to study the merging during cycling with 3.75 MV/cm at different temperature again ranging from 298 K to 136 K. Similar to Figure 3, the cycling time necessary to re-merge the peak completely increases as the temperature decreases (see Figure S6 and Figure S8). The corresponding Arrhenius-plots (Figure S7 and Figure S9) reveal activation energies in the range of 100 meV – 200 meV, which is slightly higher than that for wake-up and fatigue but, again, frequency dependent as shown in Table 1. EA is expected to be a function of the field amplitude21 because this amplitude bends the energetic potential and therefore, the barrier between the opposite stable ion positions. Due to the limitations given by early dielectric breakdown at too high fields and non-saturated hystereses at too low fields, we only used 3.75 MV/cm as the wake-up/re-merging amplitude. However, similar trends could be achieved by varying the split-up phenomenon relative to the fixed remerging amplitude. Here, split-up fields of 1.9 MV/cm and 2.3 MV/cm were used. As already stated with respects to the split-up experiments shown Figure 1, more cycles are necessary to remerge a split-up induced by a higher subcycling field. As can be seen from Table 1, a higher subcycling amplitude results in a lower frequency dependence and an overall lower activation energy. The former is in total agreement of the findings in ref. 21, which reports higher activation energies for higher merging amplitudes, whereas we encounter the same for lower split-up amplitudes. The latter, i.e. the changed frequency dependence, supports the assumption of different mechanisms being involved in all of the phenomena studied here. Again, a difference in I-E and P-E characteristics is observed, that might account for the different values observed

ACS Paragon Plus Environment

10

Page 11 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

for different split-up amplitudes and frequencies (see sections S5 and S7 of the Supporting Information). Compared to the values of around 0.6 eV to 1.2 eV obtained for the wake-up effects (aging/deaging) in PZT14,21,24 or BaTiO325, all activation energies are nearly one order of magnitude lower. Also the temperature range necessary to study the phenomena on a reasonably accessible time scale is different. In these perovskites, the oxygen vacancies are also the defects hold responsible for polarization fatigue23. HfO2 (resistive switching26) and the chemically very similar ZrO2 (fuel cells27,28) are known for their high oxygen conductivity.29,30 Calculated barrier heights for single jumps between distinct positions in the monoclinic lattice can be as low as 50 meV, but overall effective activation energies were calculated to be 0.7 eV for double positively charged O vacancies and even 2.4 eV for neutral ones.31 The apparent difference to the observations in this work are mainly attributed to the different magnitudes of the applied electric field – several MV/cm for Sr:HfO2 instead of some 10 kV/cm for PZT. The regime might be altered between the movement and formation of defects within the volume or along domain or grain boundaries. Former studies of Si:HfO2 revealed, that a domain consists of several grains (~ 30 nm lateral diameter)32, i.e. many domains contain paths that might be preferential for electronic or ionic conduction. In reliability studies on HfO2 based gate stacks, the activation energies for dielectric breakdown were also reported to increase with higher frequencies33 similar to what was observed here for fatigue. It is well-known that dielectric breakdown in HfO2 is related to the generation of oxygen vacancies near the interface layer.34 In PZT based capacitors, the fatigue issues were also related to a creation of O vacancies. The use of Ir or IrO2 electrodes solved this problem.23 Thus, it would not be surprising if future studies revealed a dominant role of oxygen vacancies for the defect related phenomena of HfO2 based ferroelectrics presented

ACS Paragon Plus Environment

11

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 25

here. Although this picture might seem consistent at the present stage of the studies, another possibility to be considered are (field induced) phase transitions. Just to give some examples to assess the order of magnitude: Clima et al.2 reported values between 230 meV and 630 meV for the switching barrier between the two opposite polarization states, which depended on the dopant material. Reyes-Lillo et al.5 calculated a barrier of 35 meV for the transition from tetragonal (P42/nmc, non-polar, paraelectric) to the orthorhombic (Pbc21, polar, ferroelectric) phase in ZrO2. An upper limit for the transition between the two paraelectric phases of monoclinic (P21/c, non-polar) and tetragonal symmetry was simulated to be ~ 200 meV for both HfO2 and ZrO2.9 However, in the following only the explanations related to defect movement are considered since they are established for fatigue and wake-up in conventional ferroelectric such as PZT.

Table 1. Comparison of activation energies for wake-up, fatigue and re-merging after split-up. phenomenonA

f = 100 Hz

f = 1000 Hz

f = 10000 Hz

84(±6) / 88(±7)

98(±20) / 99(±24)

105(±12) / 126(±22)

48(±7)

56(±14)

103(±13)

re-merging after split-up at 1.9 V: EA,7 / EA,9 in meV

118(±38) / 132(±13)

142(±117) / 152(±37)

165(±37) / 185(±26)

re-merging after split-up at 2.3 V: EA,7 / EA,9 in meV

105(±29) / 129(±22)

111(±40) / 129(±28)

124(±28) / 138(±35)

wake-up: EA,7 / EA,9 in meV fatigue: EA in meV

To especially explain the analogue subcycling dependent split-up phenomenon, a model based on the movement of defects during the switching processes is discussed. The main assumption is:

A

EA,7 and EA,9 are the activation energies for the 7th and 9th harmonic, respectively. For fatigue, only one value is given, since the maximum in Pr was used as criterion.

ACS Paragon Plus Environment

12

Page 13 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

Static, i.e. non-switching regions are a preferential location for defects to accumulate since such an environment is more stable – both from the structural and electrical point of view. Figure 4 shows how the Gibbs energy vs. (switching) ion position is related to the electric field where the switching current peaks are observed. A higher barrier between the two minima in Gibbs energy results in a higher switching field for the corresponding ion. If this accumulation of defects is expected to take place near domain boundaries, electrodes or within the volume is not specified here. The increasing number of defects progressively impedes the switching of the corresponding domain similar to what is discussed for the mechanisms behind the polarization fatigue: Defects either pin domain walls or inhibit the seeds of domains with opposite polarization near the electrodes.35 It should be noted, that the obtained activation energies are not to be confused with the barrier between to two stable stats in the Gibbs energy vs. ion position shown in the model. The EA values given here, describe the process of defect movement, which is proposed to be the origin of a change in the local switching barriers. From the broad current peaks in Figure 1 it is clear that the switching field is distributed over a wide field range and lowering the cycling field automatically results in not switching the regions with higher coercive fields any longer. Cycling again with the higher field, these formerly static domains appear as a split-up peak on the higher field side whereas the formerly cycled regions are represented by a peak located below the cycling field. During cycling, the split-up becomes more pronounced. So, the switching of the static domains requires higher and higher fields, whereas polarization reversal of the switched regions becomes more and more easy. Exactly this is attributed to the distribution of the defects in the film. Hence, this model is able to explain all experimental observation related to the split-up of multiple peaks in any desired configuration as shown in the following. Moreover, it automatically implies that the situation before the wake-up

ACS Paragon Plus Environment

13

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 25

is a result of a somehow processing related non-uniform defect distribution in addition to a coercive field distribution caused by e.g. different grain sizes that appears after the wake-up.

Figure 4. Static domains are assumed to be a preferential place for defect accumulation. Further assuming defect related domain wall pinning or and seed inhibition mechanism as they are known from ferroelectric fatigue, a progressive inhibition of the non-switched regions and a facilitated switching of the regions with lower switching fields is concluded. The expected change in Gibbs energy vs. ion position and I-E curve are sketched on the right.

Given the lack of deeper understanding in the physical mechanism involved in the novel splitup phenomenon, it is helpful to at least proof, if the proposed model is reasonable from a mathematical point of view. According to the above experimental data, Sr:HfO2 exhibits a capacitive-type response, which depends upon the previous history of voltages applied to this material. In modern parlance,36 such a system is classified as a voltage-controlled memcapacitive

ACS Paragon Plus Environment

14

Page 15 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

(memory capacitive) system. Voltage-controlled memcapacitive systems37 are described by the equations

q ( t ) = C ( x,V , t ) V ( t ) , dx = f ( x, V , t ) , dt

(2) (3)

where q(t) is the charge on the capacitor at time t, V(t) is the applied voltage, C is the memcapacitance, which depends on the state of the system and can vary in time, x is a set of n state variables describing the internal state of the system, and f is a continuous n-dimensional vector function. An important step in the memory device modeling is the identification of specific physical mechanisms

38

responsible for memory response and associated internal state

variables. In the Supporting Information S8 we present a network model of memory effects in Sr:HfO2, in which a discretized density of defects plays the role of internal state variables. This model is able to reproduce all properties of the split-up/merging phenomenon as shown in Figure S11.

CONCLUSION A more detailed look into defect related mechanisms of a rather new, promising, ferroelectric material was presented. As a simple binary oxide, HfO2 could serve as model systems to further study the fundamentals of ferroelectricity in a system without complex composition and chemistry. A time saving and convenient approach harmonic analysis was used to study the temperature and frequency dependence of three phenomena in Sr:HfO2: 1) wake-up, 2) polarization fatigue and 3) pinching/depinching of the hysteresis by electric field cycling. The obtained activation energies were in the range of 50 meV to 200 meV for wake-up, fatigue and the split-up/merging effect, but showed different frequency dependences. Compared to the

ACS Paragon Plus Environment

15

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 25

values obtained for PZT, these energies are nearly one order of magnitude lower, which was attributed to the comparably high applied fields for Sr:HfO2. Also the much higher number of grain boundaries and corresponding defects per volume compared to the PZT ceramics in other studies should be taken into account. Moreover, the novel analog split-up/merging phenomenon was, to the author’s knowledge, reported here for the first time. A modelling approach based on memcapacitive behavior was used to explain the observed phenomena as a result of different local defect concentrations. It was assumed that the observed cycling response is associated with a migration of defects from actively switching domains (at a given cycling field amplitude) to passive ones. Although temperature dependent studies were only performed for 3.4 cat% Sr, the behavior is similar for other concentrations and other dopant materials. Tuning the kinetics of this effect by appropriate processing, it might be of special interest for multilevel cell memories39 or smart operating schemes with a rejuvenation of a fatigued Pr by only a few pulses of higher amplitude after a certain amount of stress cycles to prolong the endurance of the memory cell.

ACS Paragon Plus Environment

16

Page 17 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

EXPERIMENTAL PROCEDURES Metal-ferroelectric-metal (MFM) stacks were fabricated on a lightly p-doped Si substrate. TiN bottom electrode and the Sr:HfO2 dielectric, 10 nm each, were deposited by reactive magnetron sputtering at room temperature and thermal atomic layer deposition (ALD) at 300 °C, respectively. The Sr content of the sample was adjusted by the ratio of HfO2 and SrO ALD cycles using a HfCl4/H2O and a Sr(tBu3Cp)2/H2O process. The resulting Sr content was determined to be 3.4 cat% (cat% = Sr/[Hf + Sr]) by Rutherford backscattering spectroscopy. The further sample preparation included a reactive magnetron sputtering of the 12 nm TiN top electrodes (TE) at 200 °C and a N2 anneal for 20 s at 800 °C to crystallize the previously amorphous Sr:HfO2. Finally, circular contact pads consisting of a 10 nm adhesion layer of Ti and 50 nm Pt were deposited through a shadow mask on top of the samples in an electron beam evaporator. These pads define the MFM capacitor structures subsequently patterned by an SC1 etch to remove the TiN TE. Electrical measurements were performed with a TF Analyzer 3000 with FE-Module (aixACCT Systems). Harmonic analysis20,21 was applied to study wake-up and the split-up/merging effect. Phase jumps in 7th and 9th harmonic were used as indicator for a transition between a pinched and a depinched hysteresis shape. In contrast to the direct measurement with lock-in technique presented in references 20 and 21, here the “Fatigue Measurement” environment of the TF Analyzer 3000 was used. Within this environment, rectangular stress cycles and triangular measurement cycles were applied during the dynamic hysteresis measurements to extract a P-E hysteresis three times per decade of stress cycles. Afterwards, the data was subjected to a Fourier series expansion to obtain amplitude and phase values of the respective harmonics and to plot this data against the number of field cycles as shown in Figure 3 a). This procedure was repeated for different temperatures to plot the time of

ACS Paragon Plus Environment

17

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 18 of 25

the phase jump versus inverse temperature to extract the activation energy of 7th and 9th harmonic. For the fatigue effect, the approach was different. Here, the onset of fatigue was linked to point in time, when Pr started decreasing and not to phase jumps in any harmonic.

ASSOCIATED CONTENT Supporting Information. Sample preparation, measurement setup. Harmonic analysis and a comparison to the Preisach model. Polarization hystereses and transient current graphs for different temperatures. Arrhenius-plots for all extracted activation energies. Details on the memcapacitive network model and its results. This material is available free of charge via the Internet at http://pubs.acs.org. AUTHOR INFORMATION Corresponding Author * E-mail: [email protected] ACKNOWLEDGMENT Bernd Reichenberg from aixACCT Systems is gratefully acknowledged for the valuable discussions and rapid development of the upgrade necessary to perform the measurements for this work. Moreover, the German Research Foundation (Deutsche Forschungsgemeinschaft) is acknowledged for funding this research within the project “Inferox” (MI 1247/11-1).

ACS Paragon Plus Environment

18

Page 19 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

REFERENCES 1.

Böscke, T. S.; Müller, J.; Bräuhaus, D.; Schröder, U.; Böttger, U. Ferroelectricity in

Hafnium Oxide Thin Films. Appl. Phys. Lett. 2011, 99, 102903. 2.

Clima, S.; Wouters, D. J.; Adelmann, C.; Schenk, T.; Schroeder, U.; Jurczak, M.;

Pourtois, G. Identification of the Ferroelectric Switching Process and Dopant-Dependent Switching Properties in Orthorhombic HfO2: A First Principles Insight. Appl. Phys. Lett. 2014, 104, 092906. 3.

Huan, T. D.; Sharma, V.; Rossetti Jr., G. A.; Ramprasad, R. Pathways Towards

Ferroelectricity in Hafnia. Phys. Rev. B 2014, 90, 064111 4.

Bennett, J. W.; Garrity, K. F.; Rabe, K. M.; Vanderbilt, D. Orthorhombic ABC

Semiconductors as Antiferroelectrics. Phys. Rev. Lett. 2013, 110, 017603. 5.

Reyes-Lillo, S. E.; Garrity, K. F.; Rabe, K. M. Antiferroelectricity in Thin Film ZrO2

from First Principles. Phys. Rev. B. 2014, 90, 140103(R). 6.

Park, M. H.; Kim, H. J.; Kim, Y. J.; Lee, W.; Kim, H. K.; Hwang, C. S. Effect of

Forming Gas Annealing on the Ferroelectric Properties of Hf0.5Zr0.5O2 Thin Films with and without Pt Electrodes. Appl. Phys. Lett. 2013, 102, 112914. 7.

Lomenzo, P. D.; Zhao, P.; Takmeel, Q.; Moghaddam, S.; Nishida, T.; Nelson, M.;

Fancher, C. M.; Grimley, E. D.; Sang, X.; LeBeau, J. M.; Jones, J. L. Ferroelectric Phenomena in Si-Doped HfO2 Thin Films with TiN and Ir Electrodes. J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.: Mater., Process., Meas., Phenom. 2014, 32, 03D123.

ACS Paragon Plus Environment

19

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

8.

Page 20 of 25

Park, M. H.; Kim, H. J.; Kim, Y.J.; Moon, T.; Kim, K.D.; Hwang, C. S. Thin HfxZr1-xO2

Films: A New Lead-Free System for Electrostatic Supercapacitors with Large Energy Storage Density and Robust Thermal Stability. Adv. Energy Mater. 2014, 1400610, 1–7. 9.

Luo, X.; Zhou, W.; Ushakov, S. V.; Navrotsky, A.; Demkov, A. A. Monoclinic to

Tetragonal Transformations in Hafnia and Zirconia: A Combined Calorimetric and Density Functional Study. Phys. Rev. B 2009, 80, 134119. 10.

Bohr, M.; Chau, R.; Ghani, T.; Mistry, K. The High-k Solution. IEEE Spectrum 2007, 44,

29–35. 11.

International Technology Roadmap for Semiconductors – Emerging Research

Devices.2013. http://www.itrs.net/Links/2013ITRS/2013Chapters/2013ERD.pdf (accessed Sep. 28, 2014) 12.

Müller, J.; Yurchuk, E.; Schlösser, T.; Paul, J.; Hoffmann, R.; Mueller, S.; Martin, D.;

Slesazeck, S.; Polakowski, P.; Sundqvist, J.; Czernohorsky, M.; Seidel, K.; Kücher, P.; Boschke, R.; Trentzsch, M.; Gebauer, K.; Schröder, U.; Mikolajick, T. “Ferroelectricity in HfO2 Enables Nonvolatile Data Storage in 28 nm HKMG”. Symposium on VLSI Technology (VLSIT), June, 2012. 13.

Schenk, T.; Mueller, S.; Schroeder, U.; Materlik, R.; Kersch, A.; Popovici, M.;

Adelmann, C.; Van Elshocht, S.; Mikolajick, T. Strontium Doped Hafnium Oxide Thin Films: Wide Process Window for Ferroelectric Memories. ESSDERC 2013, Proc. Eur. Solid-State Device Res. Conf., 43rd 2013, 260–263.

ACS Paragon Plus Environment

20

Page 21 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

14.

Carl, K.; Hardtl, K. H. Electrical After-Effects in Pb(Ti, Zr)O3 Ceramics. Ferroelectrics

1977, 17, 473–486. 15.

Schenk, T.; Yurchuk, E.; Müller, S.; Schröder, U.; Starschich, S.; Böttger, U.;

Mikolajick, T. Physical Roots for the Deformation of Ferroelectric Hystereses. Appl. Phys. Rev in Review. 16.

Olsen, T.; Schroder, U.; Muller, S.; Krause, A.; Martin, D.; Singh, A.; Müller, J.; Geidel,

M.; Mikolajick, T. Co-Sputtering Yttrium into Hafnium Oxide Thin Films to Produce Ferroelectric Properties. Appl. Phys. Lett. 2012, 101, 082905. 17.

Zhou, D.; Xu, J.; Li, Q.; Guan, Y.; Cao, F.; Dong, X.; Müller, J.; Schenk T.; Schröder, U.

Wake-up Effects in Si-Doped Hafnium Oxide Ferroelectric Thin Films. Appl. Phys. Lett. 2013, 103, 192904. 18.

Schroeder, U.; Yurchuk, E.; Müller, J.; Martin, D.; Schenk, T.; Polakowski, P.;

Adelmann, C.; Popovici, M.; Kalinin, S. V.; Mikolajick, T. Impact of Different Dopants on the Switching Properties of Ferroelectric Hafniumoxide. Jpn. J. Appl. Phys. 2014, 53, 08LE02. 19.

Menou, N.; Muller, C.; Baturin, I. S.; Shur, V. Y.; Hodeau, J.-L. Polarization Fatigue in

PbZr0.45Ti0.55O3-Based Capacitors Studied from High Resolution Synchrotron X-Ray Diffraction. J. Appl. Phys. 2005, 97, 064108. 20.

Morozov, M. I. Softening and Hardening Transitions in Ferroelectric Pb(Zr,Ti)O3

Ceramics. PhD thesis, EPFL – Swiss Federal Institute of Technology, Lausanne, Nov., 2005

ACS Paragon Plus Environment

21

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

21.

Page 22 of 25

Morozov, M. I.; Damjanovic, D. Hardening-Softening Transition in Fe-Doped

Pb(Zr,Ti)O3 Ceramics and Evolution of the Third Harmonic of the Polarization Response. J. Appl. Phys. 2008, 104, 034107 22.

Preisach, F. Über die Magnetische Nachwirkung. Z. Physik 1935, 94, 277–302.

23.

Scott, J. F., Ferroelectric Memories, Springer-Verlag, Berlin. 2000

24.

Robels, U.; Schneider-Störmann, L.; Arlt, G. Dielectric Aging and its Temperature

Dependence in Ferroelectric Ceramics. Ferroelectrics 1995, 168, 301–311. 25.

Warren, W. L.; Vanheusden, K.; Dimos, D.; Pike, G. E.; Tuttle, B. A. Oxygen Vacancy

Motion in Perovskite Oxides. J. Am. Ceram. Soc. 1996, 79, 536–538. 26.

Goux, L.; Czarnecki, P.; Chen, Y. Y.; Pantisano, L.; Wang, X. P.; Degraeve, R.;

Govoreanu, B.; Jurczak, M.; Wouters, D. J.; Altimime, L. Evidences of Oxygen-Mediated Resistive-Switching Mechanism in TiN\HfO2\Pt Cells. Appl. Phys. Lett. 2010, 97, 243509. 27.

Tucker, M. C. Progress in Metal-Supported Solid Oxide Fuel Cells: A Review. J. Power

Sources 2010, 195, 4570–4582. 28.

Laguna-Bercero, M.A. Recent Advances in High Temperature Electrolysis Using Solid

Oxide Fuel Cells: A Review. J. Power Sources 2012, 203, 4–16. 29.

Kharton, V. V.; Yaremchenko, A. A.; Naumovich, E. N.; Marques, F. M. B. Research on

the Electrochemistry of Oxygen Ion Conductors in the Former Soviet Union III. HfO2-, CeO2and ThO2-Based Oxides. J. Solid State Electrochem. 2000, 4, 243–266.

ACS Paragon Plus Environment

22

Page 23 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

30.

Kharton, V. V.; Naumovich, E. N.; Vecher, A. A. Research on the Electrochemistry of

Oxygen Ion Conductors in the Former Soviet Union I. ZrO2-Based Ceramic Materials. J. Solid State Electrochem. 1999, 3, 61–81. 31.

Capron, N.; Broqvist, P.; Pasquarello, A. Migration of oxygen vacancy in HfO2 and

across the HfO2/SiO2 interface: A first-principles investigation. Appl. Phys. Lett. 2007, 91, 192905. 32.

Martin, D.; Müller, J.; Schenk, T.; Arruda, T. M.; Kumar, A.; Strelcov, E.; Yurchuk, E.;

Müller, S.; Pohl, D.; Kersch, A.; Schröder, U.; Kalinin, S. V.; Mikolajick, T. Ferroelectricity in Si-Doped HfO2 Revealed: A Binary Lead-Free Ferroelectric. Adv. Mater. 2014, DOI: 10.1002/adma.201403115. 33.

Knebel, S.; Kupke, S.; Schroeder, U.; Slesazeck, S.; Mikolajick, T.; Agaiby, R.;

Trentzsch, M. Influence of Frequency Dependent Time to Breakdown on High-k/Metal Gate Reliability. IEEE Trans. Electron Devices 2013, 60, 2368–2371. 34.

Bersuker, G.; Heh, D.; Young, C.; Park, H.; Khanal, P.; Larcher, L.; Padovani, A.;

Lenahan, P.; Ryan, J.; Lee, B. H.; Tseng, H.; Jammy, R. Breakdown in the Metal/High-k Gate Stack: Identifying the ‘Weak Link’ in the Multilayer Dielectric. Proceedings of the IEEE International Electron Devices Meeting 2008; 1–4. 35.

Tagantsev, A. K.; Stolichnov, I.; Colla, E. L.; Setter, N. Polarization Fatigue in

Ferroelectric Films: Basic Experimental Findings, Phenomenological Scenarios, and Microscopic Features. J. Appl. Phys. 2001, 90, 1387.

ACS Paragon Plus Environment

23

ACS Applied Materials & Interfaces

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

36.

Page 24 of 25

Di Ventra, M.; Pershin, Y. V. Memory Materials: A Unifying Description. Mater. Today

2011, 14, 584-591. 37.

Di Ventra, M. Pershin, Y. V. Chua, L.O. Circuit Elements With Memory: Memristors,

Memcapacitors, and Meminductors. Proc. IEEE 2009, 97, 1717–1724. 38.

Pershin, Y. V.; Di Ventra, M. Memory Effects in Complex Materials and Nanoscale

Systems. Adv. Phys. 2011, 60, 145–227. 39.

Bauer, M.; Alexis, R.; Atwood, G.; Baltar, B.; Fazio, A.; Frary, K.; Hensel, M.; Ishac,

M.; Javanifard, J.; Landgraf, M.; Leak, D.; Loe, K.; Mills, D.; Ruby, P.; Rozman, R.; Sweha, S.; Talreja, S.; Wojciechowski, K. A Multilevel-Cell 32 Mb Flash Memory. IEEE International Solid-State Circuits Conference (ISSCC), Digest of Technical Papers 1995; 132–133

ACS Paragon Plus Environment

24

Page 25 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Applied Materials & Interfaces

TABLE OF CONTENS GRAPHIC

ACS Paragon Plus Environment

25